New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
TopNode not defined #22
Comments
Which version of Julia are you using? On Fri, Sep 2, 2016 at 5:13 PM Yao Lu notifications@github.com wrote:
|
The current release of Knet is not yet compatible with 0.5. I will release On Fri, Sep 2, 2016 at 6:58 PM Yao Lu notifications@github.com wrote:
|
Hi, We are testing the Julia port in PowerPC machines. Unfortunately, Julia compiles only in 0.5 Julia. Viral is asking us to test the port and I'm interested to test the GPU support of Julia using deep learning package like Knet. I have similar error above during Knet's building the CUDArt. It will be cool if Knet runs in PowerPC: using Knet |
I will take a look...
|
I switched the latest master to 0.8 development. You should be able to julia> Pkg.checkout("Knet") Let me know if this solves your problem. Knet8 is under development, the documentation and updates are not up to best, On Sat, Sep 3, 2016 at 11:00 PM Deniz Yuret denizyuret@gmail.com wrote:
|
Thanks. I tried running in 0.5 but CUDArt has issues. Viral sent us 0.6.0 for testing. I did Pkg.build("Knet") after checkout and it compiled successfully in 0.6 Julia. gpu(true) also worked but running MNIST.main() result to errors: it may indicate I'll convert Float32 to Float64? opts=(:seed,-1)(:batchsize,100)(:hidden,Int64[])(:epochs,10)(:lr,0.5)(:atype,"KnetArray{Float32}")(:winit,0.1)(:fast,false) |
It does not look like it is able to locate the libknet8 library. Is it in deniz On Tue, Sep 6, 2016 at 4:55 PM Paulito Palmes notifications@github.com
|
yes, libknet8.so and libknet.so are in the Pkg.dir("Knet/src") the value of Knet.libknet8 is "" |
sorry, still not working. I thought compiling files in src7 will help but it did not work. it compiled successfully but MNIST.main() still has similar errors described above. |
src7 is deprecated, I am only keeping it around temporarily, no need to On Tue, Sep 6, 2016 at 5:18 PM Paulito Palmes notifications@github.com
|
is there a manual way to make sure libknet8 will be available to Knet Julia? the libs are there so I think we just have to export the location? |
Normally the info under Knet/deps should cause the compilation libknet8 On Tue, Sep 6, 2016 at 5:29 PM Paulito Palmes notifications@github.com
|
i just followed the steps: Pkg.checkout("Knet") |
Hmm, I should test with 0.6 as well. Is there a downloadable version On Tue, Sep 6, 2016 at 5:47 PM Paulito Palmes notifications@github.com
|
let me send the link. viral sent us the link but i'm afraid he sent us binary version for powerpc machines. let me check the email... |
here's a snapshot of the email from Viral. I think the first link is the binary version for ppc while the second link is the source code. Can you try this one? I believe this error is addressed now. Also, Elliot enabled the nightlies: |
One of my students had the same problem -- apparently the precompiled On Tue, Sep 6, 2016 at 6:45 PM Paulito Palmes notifications@github.com
|
thanks. I removed Knet.ji but got different error: julia> using Knet |
Can you do a Pkg.update(), I don't think you have the latest AutoGrad On Tue, Sep 6, 2016 at 8:43 PM Paulito Palmes notifications@github.com
|
MNIST.main() with gpu(true) works now with the latest AutoGrad. I timed gpu(false) vs gpu(true) for MNIST.main(). They finished both around 3 to 4 seconds. It feels no difference between the two. Even the printing of results look similar speed. By the way, latest Knet no longer uses compile and other macros. Is there a documentation for the latest Knet? Thanks. |
I am writing the documentation as we speak!!! In summary, Knet8 does not compile models any more, it uses AutoGrad to When I test with MNIST.main("--fast") I get 2.82 for the cpu and 2.12 for best, On Tue, Sep 6, 2016 at 9:04 PM Paulito Palmes notifications@github.com
|
thanks for the explanation. I'm just trying to make sure that GPU is really working ;). it is great to have Knet GPU working now in Power. So far, this is the only deep learning package easy to install relative to other deep learning packages I tried and I very much prefer the programming logic of Knet. the documentation of the previous version is great and really looking forward to use Knet in my workflow. thanks for the great package!!! Sent from my iPhone
|
You are very welcome. You can sign up on the knet-users mailing list (very BTW, I couldn't compile Julia 6 on the intel servers I have (Centos 6.5) deniz On Tue, Sep 6, 2016 at 11:40 PM Paulito Palmes notifications@github.com
|
I'll ask Viral if he has Intel binaries also. I like to contribute someday on the aspects of running Knet in PowerPC machines. I like the simplicity of Knet which is true to the KISS principles of UNIX ;) Sent from my iPhone
|
by the way, the latest version of Knet's REQUIRE commented out CUDArt, ArgParse, and GZip which are needed to run MNIST.main() |
Yes, I decided to minimize the requirements to run everything under src and On Wed, Sep 7, 2016 at 5:10 PM Paulito Palmes notifications@github.com
|
@denizyuret Thank you! Knet8 is usable for me. |
knet7 will be maintained with bugfixes, future work will focus on knet8. On Thu, Sep 8, 2016, 07:18 Yao Lu notifications@github.com wrote:
|
julia> using Knet
WARNING: Base.SparseMatrix is deprecated.
likely near /home/rluser/.julia/v0.5/CUSPARSE/src/CUSPARSE.jl:2
WARNING: Method definition (::Type{Knet._CudaArray})(CUDArt.CudaArray{#T<:Any, #N<:Any}) in module Knet at /home/rluser/.julia/v0.5/Knet/src/util/cudart.jl:127 overwritten at /home/rluser/.julia/v0.5/Knet/src/util/cudart.jl:128.
WARNING: Base.writemime is deprecated.
likely near /home/rluser/.julia/v0.5/Knet/src/util/cudart.jl:133
WARNING: Base.writemime is deprecated.
likely near /home/rluser/.julia/v0.5/Knet/src/util/cudart.jl:133
WARNING: Base.LambdaStaticData is deprecated, use LambdaInfo instead.
likely near /home/rluser/.julia/v0.5/Knet/src/util/deepcopy.jl:21
ERROR: LoadError: LoadError: UndefVarError: TopNode not defined
in include_from_node1(::String) at ./loading.jl:426 (repeats 2 times)
in eval(::Module, ::Any) at ./boot.jl:234
in require(::Symbol) at ./loading.jl:357
while loading /home/rluser/.julia/v0.5/Knet/src/util/deepcopy.jl, in expression starting on line 21
while loading /home/rluser/.julia/v0.5/Knet/src/Knet.jl, in expression starting on line 17
The text was updated successfully, but these errors were encountered: